NONOBE Koji Kyoto University, Graduate School of Informatics, Assistant Professor, 情報学研究科, 助手 (40324678)
HORIYAMA Takashi Nara Institute of Science and Technology, Graduate School of Information Science, Assistant Professor, 情報科学研究科, 助手 (60314530)
YAGIURA Mutsunori Kyoto University, Graduate School of Informatics, Lecturer, 情報学研究科, 講師 (10263120)
SHINANO Yuji Tokyo University of Agriculture and Technology, Faculty of Technology, Lecturer, 工学部, 講師 (00297623)
|Budget Amount *help
¥10,800,000 (Direct Cost : ¥10,800,000)
Fiscal Year 2000 : ¥2,900,000 (Direct Cost : ¥2,900,000)
Fiscal Year 1999 : ¥4,100,000 (Direct Cost : ¥4,100,000)
Fiscal Year 1998 : ¥3,800,000 (Direct Cost : ¥3,800,000)
Recently, many types of metaheuristics based on local search, e. g., simulated annealing, genetic algorithm, and tabu search, have been proposed, and successfully applied to various combinatorial optimization problems. The objective of this research is to develop powerful general-purpose algorithms for combinatorial problems arising in real applications. To achieve this purpose, we developed metaheuristics algorithms for the following problems : (a) Weighted Constraint Satisfaction Problem, (b) Generalized Assignment Problem, (c) Resource Constrained Project Scheduling Problem, (d) Set Covering Problem, (e) Cutting Stock Problem, (f) Vehicle Routing Problem, (g) Rectangle Packing Problem, and others. These problems are representative combinatorial optimization problems, and their conventional formulations are rather simple. In our research, aiming at improving the applicability of algorithms, we extended these formulations so that more practical and complicated problems can be handled.
In designing algorithms, we adopted iterated local search and tabu search as the frameworks, and defined search space and neighborhoods carefully so as to attain better performance. We also incorporated some mechanisms that adaptively control program parameters during the search, which can reduce users' efforts to tune the parameters appropriately.
In computational experiments, we solved many benchmark instances, and succeeded to improve the best known values for many instances. As to practical problems, our codes could also find solutions of high quality in reasonable computational time. Some codes are already being used in practical applications.